Document Sample

Banco Central de Chile Documentos de Trabajo Central Bank of Chile Working Papers N° 516 Abril 2009 INTERBANK RATE AND THE LIQUIDITY OF THE MARKET Luis A. Ahumada Álvaro García Luis Opazo Jorge Selaive La serie de Documentos de Trabajo en versión PDF puede obtenerse gratis en la dirección electrónica: http://www.bcentral.cl/esp/estpub/estudios/dtbc. Existe la posibilidad de solicitar una copia impresa con un costo de $500 si es dentro de Chile y US$12 si es para fuera de Chile. Las solicitudes se pueden hacer por fax: (56-2) 6702231 o a través de correo electrónico: bcch@bcentral.cl. Working Papers in PDF format can be downloaded free of charge from: http://www.bcentral.cl/eng/stdpub/studies/workingpaper. Printed versions can be ordered individually for US$12 per copy (for orders inside Chile the charge is Ch$500.) Orders can be placed by fax: (56-2) 6702231 or e-mail: bcch@bcentral.cl. BANCO CENTRAL DE CHILE CENTRAL BANK OF CHILE La serie Documentos de Trabajo es una publicación del Banco Central de Chile que divulga los trabajos de investigación económica realizados por profesionales de esta institución o encargados por ella a terceros. El objetivo de la serie es aportar al debate temas relevantes y presentar nuevos enfoques en el análisis de los mismos. La difusión de los Documentos de Trabajo sólo intenta facilitar el intercambio de ideas y dar a conocer investigaciones, con carácter preliminar, para su discusión y comentarios. La publicación de los Documentos de Trabajo no está sujeta a la aprobación previa de los miembros del Consejo del Banco Central de Chile. Tanto el contenido de los Documentos de Trabajo como también los análisis y conclusiones que de ellos se deriven, son de exclusiva responsabilidad de su o sus autores y no reflejan necesariamente la opinión del Banco Central de Chile o de sus Consejeros. The Working Papers series of the Central Bank of Chile disseminates economic research conducted by Central Bank staff or third parties under the sponsorship of the Bank. The purpose of the series is to contribute to the discussion of relevant issues and develop new analytical or empirical approaches in their analyses. The only aim of the Working Papers is to disseminate preliminary research for its discussion and comments. Publication of Working Papers is not subject to previous approval by the members of the Board of the Central Bank. The views and conclusions presented in the papers are exclusively those of the author(s) and do not necessarily reflect the position of the Central Bank of Chile or of the Board members. Documentos de Trabajo del Banco Central de Chile Working Papers of the Central Bank of Chile Agustinas 1180 Teléfono: (56-2) 6702475; Fax: (56-2) 6702231 Documento de Trabajo Working Paper N° 516 N° 516 INTERBANK RATE AND THE LIQUIDITY OF THE MARKET Luis A. Ahumada Álvaro García Luis Opazo Jorge Selaive Banco Central de Chile Banco Central de Chile Banco Central de Chile Banco de Crédito e Inversiones Resumen En este trabajo estudiamos la dinámica diaria de la tasa interbancaria en Chile, con especial atención al rol de la liquidez provista a través de depósitos bancarios y operaciones de mercado abierto realizadas por el Banco Central. El principal objetivo de este trabajo es evaluar la información contenida en datos desagregados y de alta frecuencia de esas variables. Los principales resultados se encuentran relacionados a la significancia económica de la velocidad de convergencia, los efectos calendario y de las operaciones con pacto de retroventa (repo). El Banco Central juega un rol más importante drenando que inyectando liquidez por medio de operaciones monetarias discrecionales. Sin embargo, no hay asimetrías en términos de la efectividad de las inyecciones y drenajes discrecionales dependiendo de la liquidez del mercado. Adicionalmente, los bancos de mayor tamaño son menos receptivos a las operaciones monetarias, mientras que los bancos pequeños son los más sensibles a dichas operaciones, lo cual es consistente con el tradicional rol de estas categorías de bancos en la provisión de liquidez en el mercado interbancario. Finalmente, los depositantes privados no juegan un rol importante en la dinámica de la tasa interbancaria durante el período muestral. Abstract In this paper we study the dynamics of the interbank rate in Chile, with special attention to the role of liquidity provided by private depositors and by the central bank’s open market operations on a daily basis. The main aim of this paper is the use of disaggregated and high frequency data on such variables. The most relevant findings are related to the statistical and economic significance of speed of convergence, calendar effects and repos operations. The Central Bank plays a more important role injecting than draining liquidity through discretionary operations. However, there are not asymmetries in terms of the effectiveness of the discretionary injections and drainages operations depending on the liquidity market status. In terms of effect by class of bank, large- and medium-size banks are less receptive to monetary operations; by contrast small-size banks are the most responsive, which is consistent with its traditional position as a liquidity demander. Finally, private deposits do not play an important role on the dynamics of the interbank rate during the sample period. _______________ We thank Felipe Alarcón, Matías Bernier and Jorge Pérez for helpful suggestions, participants at internal seminars in Central Bank, and Carmen G. Silva who collaborated in a preliminary version of this work. We are also grateful to the Department of Open Market Operations for sharing with us the data used in this work. All errors are ours. E-mail: lahumada@bcentral.cl; agarcia@bcentral.cl; lopazo@bcentral.cl; jselaiv@bci.cl. 1. Introduction The interbank money market rate (ir) stands at the shortest end of the yield curve, and is the operational target for the monetary policy rate (mpr). Therefore, understanding the factors behind the dynamics of the ir is relevant not only for participants in the interbank market, but also for private investors and monetary authorities. Indeed, the ir is a key benchmark for interest rates in the short-term money market and its movements may have effects on the whole term structure (Taylor and Williams, 2008 among others). Moreover, the interbank market represents the first stage of the monetary transmission channel, where monetary policy actions first come into contact with the rest of the financial system. An effective monetary policy requires that the overnight interest rate remains “at an average of around” the mpr. 1 In this paper we study the dynamics of the ir in Chile, giving special attention to the role of the liquidity provided by private depositors and by the Central Bank’s open market operations. Our paper extends the previous literature mainly in three aspects. Firstly, we study the effect of liquidity provision by both the central bank and private depositors on the dynamics of the ir. To the best of our knowledge, this paper is the first attempt in the literature to broadly incorporate this dimension into the analysis of the ir. Secondly, we provide novel evidence on the behavior of the ir in an emerging economy, which is useful to fill the gap created by preceding literature mainly focusing on industrial countries.2 Finally, we take a systematic approach that involves both the time series and panel data dimensions, allowing us to have a broader picture of the factors behind the high frequency dynamics of the ir. The liquidity of the market affects directly the amount of resources that commercial banks have at their disposal and which they will consequently be willing to lend in the interbank market. However, with the exceptions of Wurtz (2004) and Hamilton (1996),3 previous empirical studies have not considered the effect of daily liquidity conditions on the analysis of the ir. In fact, the literature 1 The functioning of the Chilean interbank market is similar in structure to US and Euro area cases. Appendix 1 describes in depth technical details related to the functioning of the interbank market in Chile. 2 Evidence for the European overnight interbank rate (EONIA) and for the federal funds rate can be found in Spindt and Homeister (1988), Hamilton (1996), Balduzzi et al. (1997, 1998), Gaspar (2004), Nautz and Offermanns (2006), Prati et al. (2003) and Cocco et al. (2009), among others. 3 These authors consider the daily reserve surplus, i.e. current account holdings minus reserve requirements, as an indicator of the liquidity conditon. 1 generally analyzes the functioning of the interbank market using a general framework in which banks’ reserve positions are affected by random shocks and where the interbank market allows banks to fulfill their monthly reserve requirement (e.g., Ho and Saunders, 1985; Freixas et al, 2000; Allen and Gale, 2000). Within this scheme, an important number of empirical papers have studied the effect of periodic events affecting banks’ reserve positions. Indeed, a bulk of empirical evidence points out that variables like the last day of the reserve maintenance period, the last day of the month, the prior day to a Holy day or the day of the monetary policy meeting are relevant for explaining the daily dynamics of the ir (Hamilton, 1996; Sarno and Thornton, 2001; Prati et al., 2003; Moschitz, 2004; Nautz and Offermans, 2006; Benito et al., 2007). As mentioned above, we make use of a unique database containing detailed information about the different types of open market operations as well as deposits by depositors at the bank level. This dataset allows us to test the role of money market operations through permanent credit lines and repos. Liquidity provision by the central bank usually involves drainage and injection of funds through repo operations at mpr (discretionary operations hereafter), combined with permanent draining and injection of funds at mpr – 25 bps and mpr + 25 bps, the “floor” and “ceiling” of market rates around the target, respectively. The use of discretionary operations, instead of credit facilities, could be interpreted as a high degree of commitment of the central bank to take the ir close to the mpr, which could lead to the gap between these rates being closed faster. In this line of research, Nautz and Offermanns (2006) explore the role of the repo auction format in the Euro zone, and Prati et al. (2003) study central banks’ operating procedures and intervention styles for the Euro zone and G-7 countries. The modeling strategy for the time series follows Sarno and Thornton (2001), and Nautz and Offermanns (2006), who employ an error correction model to characterize the dynamics of the ir allowing for non-linearities.4 Given that we have high frequency data of the monetary policy operations and deposits at banks, we complement the strategies of these studies by evaluating the economic impact of different types of interventions on the short run dynamics. We also evaluate whether deposits from pension funds (PF hereafter) 4 Benito et al. (2007) follow a more statistical approach to model the EONIA. They employ several models containing jump components – for instance ARCH-Poisson-Gaussian process. 2 have a different impact on the ir from that of the others depositors. We supplement time series estimations with panel data analysis that exploits the variation of the ir across banks. The estimation strategy is a standard fixed effect panel using instrumental variables to control for potential endogeneity of some regressors. Our results indicate that the ir and the mpr move together very closely and, when these variables deviate from each other, the speed of convergence is around 30% per day. In term of the explanatory variables, the calendar effects and open market operations –especially the discretionary operations- are the most relevant in explaining the dynamics of the ir. With respect to the calendar effects, they play an important role on the dynamics of the ir –which is consistent with previous findings in this area- a situation that poses questions on which frictions drive this finding. Regarding the relevance of market liquidity provided by the central bank, we find that the central bank played an important role during the sample period, while private depositors do not help to significantly improve the explanation of the dynamics of the ir. For example, if we consider the average daily monetary operation and long-run PF’s deposits, the effect of discretionary injections, drainages and long-run PF’s deposits on the ir is 1.7, 3.4, and 0.05 bps, respectively. The permanent credit lines are not statistically significant; this situation could be due to the fact that this instrument is available on a daily frequency and, therefore, the market has internalized its operation in the valuation of the ir. The results also show a more active role of central bank injecting than draining liquidity. The effect of draining and injection operations on the ir are quite similar, but the magnitude of drainages are close to two time the injections. At the bank level, the most relevant asymmetry is due to the distinction of banks according to their size. Concretely, the large banks are less sensitive to monetary operations –which could be associated to their condition of liquidity providers- meanwhile small- size banks are the most responsive to central bank’s instruments. The rest of this paper is structured as follows. Section 2 presents the data. Section 3 performs time and panel data estimations. Section 4 concludes. 3 2. Data 2.1. Description In this paper we use three main data sets. The first consists of detailed information for each loan granted in the Chilean interbank market on a daily basis for the period of June 2006 to August 2008 (532 trading days). The data is from the Central Bank of Chile5 and include 29 banks that are active traders in the money market. It also identifies lenders and borrowers, as well as the interest rate and volume involved in each operation. With the previous information, we compute the daily ir at which banks borrow from each other as the weighted average of its operations for each day. Similarly, we construct the aggregated ir as the weighted average of the individual ir, where the weights are computed as the ratio of the volume involved in each operation to the aggregated traded volume. The other two datasets contain information with proxies of the liquidity conditions at the banking system. The first dataset comprises daily information of the operations carried out by each bank with the Central Bank of Chile (i.e., repos, auctions of promissory notes, etc.). The second dataset contains information on deposits in commercial banks grouped by PF and other investors. This last dataset was built with information provided by the Superintendencia de Bancos e Instituciones Financieras (SBIF) and the Superintendencia de Pensiones (SP). We consider several other sources of information for external and domestic variables that we include in the analysis as factors that could potentially affect the aggregated liquidity conditions. On the external side, we use the Libor-OIS spread and the VIX. On the domestic side, we consider a measure for shocks on the expected mpr, corresponding to the difference between the expected mpr derived from forward contracts and the current mpr. 2.2. The Interbank Rate and the Overnight Money Market Within the sample period, the ir has followed the policy rate closely, generally being only 1.8 bps above the mpr (see Figure 1a). In this dimension, the Central Bank has been successful in steering the short-term interest rates 5 The functioning of the Chilean interbank market is similar in structure to U.S. and European area cases. Appendix 1 describes in depth the technical details related to the functioning of the interbank market in Chile. 4 towards the mpr, in spite of the large variations of the mpr during the sample period—from 5.00 to 8.25 percent and the financial turbulences derived from the sub-prime crisis. Indeed, there are only a few episodes where the ir decouples significantly from the mpr and these episodes have been highly transitory: for instance, mid-2007 (see Figure 1b). Nonetheless, there is considerable heterogeneity of these variables across time. In fact, the average deviation during 2007 was close to 5 bps which contrasts with the 0.5 bps observed during years 2006 and 2008 (Table 1). And, in terms of volume in the interbank market, its peak in the sample occurred in 2007 (30% and 15% higher with respect to 2006 and 2008, respectively), which suggests a high appetite for liquidity coming from commercial banks at the beginning of the financial turbulences derived from the sub-prime crisis. Figure 1 Chilean Monetary Policy Rate and Interbank Rate Panel (a) Panel (b) 2006.06-2008.08 2007.07-2007.09 8,50 5,95 8,00 5,75 7,50 5,55 7,00 5,35 6,50 6,00 5,15 5,50 4,95 5,00 4,75 4,50 03-07-2007 13-07-2007 23-07-2007 02-08-2007 12-08-2007 22-08-2007 01-09-2007 11-09-2007 21-09-2007 2006,06 2006,07 2006,08 2006,10 2006,11 2007,01 2007,02 2007,04 2007,05 2007,07 2007,08 2007,10 2007,11 2008,01 2008,02 2008,03 2008,05 2008,06 2008,08 ir mpr ir mpr Source: Authors’ calculations based on information provided by the Central Bank of Chile. 5 Table 1 Interbank Market’s Descriptive Statistics (2006-2008) 2005 2006 2007 2008 2005-2008 ir (%) 3.91 5.02 5.35 6.72 5.61 ir - mpr (bps) 0.8 0.5 4.8 -0.5 1.8 Average Volume 262.8 256.7 333.5 290.9 294.8 Note: Interbank rate is expressed in percent points, the difference between the interbank and the target rate in basis points. Average interbank volume is in billions pesos. Source: Authors’ calculations based on information provided by the Central Bank of Chile. In the previously mentioned decoupling of mid-2007 (see Panel (b), Figure 1), for almost an entire month the ir was systematically above the mpr, with a spread that reached a peak of 21 bps at the beginning of September 2007. This decoupling could be related to different drivers happening simultaneously, making the identification of the incidence of these factors an extremely difficult task. First, the Chilean and international financial markets were severely hit by turbulences derived from the sub-prime crisis.6 Second, on August 9th the maximum regulatory limit on foreign assets held by PF was increased from 30 to 35 percent of the total portfolio.7 Finally, there were high expectations of an increase of the mpr on the monetary policy meeting of September 13th of that year and commercial banks seemed to anticipate a 25 bps increase in that meeting. On the quantity side, Table 2 presents some descriptive statistics of the open market operations during the sample period. A few points deserve to be mentioned. First, the average interbank operations are of similar magnitude to most of the monetary operations managed by the Central Bank.8 Second, the sample period considers the significant activity of both liquidity and draining discretionary operations, which will allow us to identify if the market responds differently to these operations. Third, if we compare the frequency of liquidity and drainages with the ir-mpr spread, it is not evident that discretionary or 6 Indeed, during August 2007 the Chilean stock market experienced a phase of unusual high volatility, falling almost 10% between the 9th and the 13th of August and fully recovering these losses in the following four trading days. 7 It was common in this period to observe Chilean money managers declaring in the current press, the possibility that pension funds could be affecting market liquidity. PF accounts on average around 20% of total deposits held by commercial banks. 8 The average of each type of operation is calculated using only the days that register each type of operation. 6 permanent operations work in a counter-cyclical fashion with respect to degree of decoupling of the ir –i.e., injections (drainages) relatively more important when the ir is above (below) the mpr. In this sense, the frequency of injections or drainages of liquidity does not seem to depend only upon the ir running above or below the mpr. This could suggest that market liquidity condition is not only measured by the ir – mpr spread. We take this situation into account in our econometric estimations by using several controls. However, and just for expositional purposes, we define a liquid market the cases when the ir<mpr and vice versa. Table 2 Descriptive Statistics of the Interbank Market (2006-2008) Average Operation Volume (in billions pesos) Variable Mean Median Maximum Interbank Operations 288.6 286.2 658.6 Discretional Injections 297.0 242.9 1,202.5 Discretional Draining 600.5 545.6 1,397.2 Permanent Liquidity Facilities 47.5 21.9 513.7 Permament Deposits Facilities 266.7 195.7 1,253.1 Illiquid Market Liquid Market Conditions Conditions Share of the Time with Positive Op Overall ( ir > mpr ) ( ir < mpr ) Interbank Operations 100% 100% 100% Injections 22% 25% 15% Draining 6% 7% 5% Permanent Liquidity Facilities 88% 89% 91% Permament Deposits Facilities 100% 100% 100% Note: Discretional Operations excludes swap operations. The average only consider days with positive operations Source: Authors’ calculations based on information provided by the Central Bank of Chile. 2.3. Bank-Level information The aggregate information allows us to understand global and domestic factors behind the dynamics of the average ir, but it hides the heterogeneity of the liquidity needs of each commercial bank. For this reason, we also explore the cross-section dimension of the data to determine whether the adjustment of the ir depends on specific characteristics of each bank. 7 To illustrate the degree of heterogeneity of the ir at the bank level, Figure 2 contains the aggregate ir – mpr spread (left panel) and the spread for each bank (right panel) from June 2006 to October 2008. It is interesting to notice that the ir exhibits significant variation across banks, covering practically all the range of +/-25 bps, with a few exceptions where the spread exceeds the ceiling of +25 bps during 2007, while the opposite occurs at the end of the sample period. Figure 2 Interbank Rate - Monetary Policy Rate Spread Panel (a) Panel (b) ir-mpr Spread: Banking System ir-mpr Spread: Level Bank 0.50 0,50 0,25 0.25 0,00 0.00 -0,25 -0.25 -0,50 -0.50 2006,06 2006,07 2006,08 2006,10 2006,11 2007,01 2007,02 2007,04 2007,05 2007,07 2007,08 2007,10 2007,11 2008,01 2008,02 2008,03 2008,05 2008,06 2008,08 2006.06 2006.06 2006.07 2006.08 2006.09 2006.10 2006.11 2006.12 2007.01 2007.02 2007.03 2007.04 2007.05 2007.05 2007.06 2007.07 2007.08 2007.09 2007.10 2007.11 2007.12 2008.01 2008.02 2008.03 2008.04 2008.05 2008.05 2008.06 2008.07 2008.08 Source: Authors’ calculations based on information provided by the Central Bank of Chile. In order to shed some light on this heterogeneity, Table 3 presents interbank market information for large-, medium- and small-size banks.9 Three facts emerge: a) the medium- and small-scale banks exhibit an asking ir approximately two times larger than large banks -1.4 vs. 3.3 and 2.5 bps, respectively;10 b) the larger-banks are willing to lend to a lower ir (2.8 bps) than medium- (3.5 bps) and small-banks (4.0 bps); and c) the size of the loans as percentage of assets is quite higher in small-scale banks (2.0%) than in large- 9 These categories of banks are the most relevant in the Chilean banking system both in terms of assets and number. More importantly, for our study, they explain more than 70% of the interbank market activity. 10 One possible explanation for the lower asking ir of large-scale banks is that these banks are able to finance their reserve needs at a lower cost. To verify if large banks have higher funding costs than medium-scale banks, we build a measure of funding costs as the ratio of monthly interest payment to total liabilities of each bank relative to the average banking system. Values higher than one of this measure reflect funding cost higher than the average. On the contrary, values lower than one reflects cheaper funding cost than the banking system. The results show that large-scale bank are able to obtain funding almost 1.2 percent cheaper than the average bank, while medium-scale banks have an average funding cost 3.5 percent higher than the average. 8 and medium-size banks -0.2% and 0.6%, respectively. These findings suggest that different class of banks participate in the interbank market for different purposes, while larger tend to use the interbank market to drain liquidity, the medium-scale participate to obtain liquidity. Table 3 Interbank Market Statistics June 2006-December 2007 Large-Scale Medium-Scale Small-Scale Average Banking Banks Banks Banks System Average Asking Spread (bps) 1,4 3,3 2,5 2,9 Average Lending Spread (bps) 2,8 3,5 4,0 2,9 Average daily Interbank Asking Volume* 25,8 14,1 5,4 10,5 % of Assets 0,2% 0,5% 2,0% 0,4% % of Financial Investment 1,7% 6,4% 8,8% 3,4% Average daily Interbank Loans Volume* 37,4 9,7 4,7 10,5 % of Assets 0,3% 0,4% 1,7% 0,4% % of Financial Investment 2,5% 4,4% 7,6% 3,4% Source: Authors’ calculations based on information provided by the Central Bank of Chile. 3. Empirical Analysis In this section we present time series and panel estimations to disentangle the main driving forces behind the dynamics of the ir, considering both global and individual liquidity needs. A time series approach is useful to analyze the role of aggregate shocks, while the panel estimation allows us to exploit the heterogeneity of the ir across banks. Jointly, we will be able to study the differential effect of shocks across different types of banks. 3.1. Time Series Analysis From a time series perspective, modeling economic variables requires evaluating if the series are stationary. Stationary variables and integrated series demand completely different modeling strategies. As Table A in Appendix 2 shows, the ir and the mpr behave very persistently during the sample period. In fact, the half-life of a shock on the ir is longer than one year, while the half-life of the mpr is even more persistent. Part of these dynamics could be explained 9 by the discrete changes in the mpr. 11 However, in order to avoid the problem of spurious results, it is necessary to test the existence of unit roots.12 We apply a battery of unit-root tests to both series, including the traditional Augmented Dickey-Fuller test, the Elliot et al. (1996) efficient test, denoted as DF-LS, and the KPSS and Phillip-Perron tests. Results successively confirm for each of these tests that it is not possible to reject the null hypothesis of a unit root for ir and the mpr series. Since the ir and the mpr move closely together and sporadically deviate from each other (see Figure 1, panel (a)), we evaluate the presence of a long-run relationship between both series. Evaluating this hypothesis is equivalent to testing whether the residual of an OLS regression between ir and mpr is non- stationary against the alternative that it is stationary. Results reject the null of unit-root for residuals, confirming the presence of a long run relationship.13 The low value for the half-life of the ir-mpr spread (less than 2 trading days) seems to confirm the stationary nature of this variable. Given the non-stationary behavior of the ir and its co-integration with the mpr, the most natural approach is an error correction model (ECM) with the mpr as the long term anchor.14 This approach is not novel in the literature. In fact, it has been applied by Nautz and Offermanns (2006), and Sarno and Thornton (2002) to model the EONIA in the Euro zone and the federal funds rate in United States, respectively. The ECM we estimate is formulated as follows: Δirt = α 0 ( irt −1 − mprt −1 ) + α1Δmprt −1 + α 2 Δirt −1 + δ ´ X + ε t , (1) where irt is the interbank rate, mprt the monetary policy rate, X other explanatory variables, and Δ the first-difference operator. The parameter α 0 is the rate at which the deviations of ir from the mpr are closed each day. The vector of other explanatory variables, X, involves several monetary operations 11 Testing for unit root in the mpr is challenging because this rate changes discretely and its increments are irregularly spaced in time. An overwhelming majority of the literature fails to reject a unit root based on the low power of unit root test when dealing with series that present infrequent changes (Hamilton and Jorda, 2002). 12 In practice, both stationary and non-stationary modeling strategies for the mpr are considered in the literature. We take one of the stands in the literature testing for the presence of unit root in the ir and the target rate. 13 Table A, third column. 14 It is worth mentioning that an error correction specification could also be obtained from a more general specification where the ir is just modeled as a function of its own lags, lags from the mpr plus other controls. 10 variables, regulatory capital requirements, institutional investor deposit variables, mpr surprises, external variables and calendar effects. Regarding monetary operations, we consider discretionary operations that provide liquidity (repos) and those that reduce it (liquidity deposits), and non discretionary instruments (permanent credit facilities) expressed in net terms, that is, liquidity injections minus drainage. It is worth mentioning that the distinction between discretional and permanent monetary operations matters for the analysis of the determinants of the ir. While the first group comprises of agreements on an occasional basis issued at the mpr, the second one corresponds to operations in which every bank is allowed to deposit (withdraw) at 25 bps below (above) the mpr. From a policy standpoint, to determine the effectiveness of discretional instruments is relevant for better fine-tuning in extraordinary episodes of decoupling of the ir from the mpr and. Regulatory capital requirements are also included since they correspond to indirect instruments used by the Central Bank to drain liquidity.15 Additionally, we consider deposits in commercial banks by PF and by other investors (i.e., insurance companies, mutual funds, households, etc.). The inclusion of PF deposits may be relevant because the share of its maintained deposits’ portfolio in whole deposits of the system is above 20 percent and, therefore, could end up influencing the ir.16 We divide PF deposits into short term and long term, corresponding to less or more than 90 days respectively. Since there is no information to classify the non-PF deposits by term, we only consider the aggregates. Other domestic variables included are monthly mpr surprises, measured as the difference between the effective mpr and the implicit expected rate in forward contracts two weeks before the monetary policy meeting. We also control for calendar effects through dummies extensively used in the literature: day of monetary policy meeting –generally the second Thursday of each month-, the day that banks must cover their reserve requirement –9th of each month-, and the day of value-added tax payment.17 Finally, in order to capture the international liquidity conditions, we consider external variables such as the 15 In Chile, banks are obligued to deposit the difference between their current liabilities and the amount equivalent to times and a half of their capital and reserves in an special account in the Central Bank. 16 PF in Chile are important players in key asset prices. For instance, Cowan, Rappaport and Selaive (2006) provide evidence of the role of PF on the exchange rate. 17 For instance, Hamilton (1996), Sarno and Thornton (2002), and Nautz and Offermanns (2006). 11 CBOE Volatility Index (VIX) and the Libor-OIS spread. While the Libor - Overnight interest swap spread capture the role of “liquidity contagion” coming from external markets, the VIX captures market expectations of near-term volatility. 3.1.1. Results The OLS estimations rely on the assumption that the independent variables are predetermined or statistically exogenous. However, it is likely that monetary operations could be endogenous to the dynamics of the ir. Central banks respond to price signals when they decide the amount they will put in the lending window in the form of repos. Similarly, when commercial banks choose to obtain funds from the interbank market over the permanent facilities alternative, they are implicitly responding to the relative cost of both sources of liquidity. Finally, given that the ir represents the shortest end of the yield curve, movements of this price could affect the amount of deposits in commercial banks. Following the previous reasoning, a straightforward OLS estimation could generate biased and inconsistent parameters. Thus, given the potential endogeneity of the covariates, we run the Hausman (1978) test to all variables. The test supports statistical exogeneity for all the variables with the exception of net permanent facilities. Therefore, we use Instrumental Variables (IV) procedure, choosing as external instruments for differences of net permanent facilities their lagged levels, lagged values of the ir-mpr spread, daily dummies and dummies for positive and negative values of the ir-mpr spread.18 A similar approach using IV estimation has also been used recently by Cocco et al (2009). Table 4 presents our estimates for the short-run dynamics of the ir, which include up to one lagged difference of the ir and the mpr. Several findings deserve attention. First, the speed of convergence of the ir to the mpr is relatively high (0.28-0.34, approximately) indicating that approximately one- third of the gap between these two variables is reduced in one day. This magnitude is a little higher but otherwise near to the estimates for the Euro interbank market by Nautz and Offermanns (2006), who find a speed of convergence of 0.26. Second, the short-run effect of changes in the mpr is 18 The set of instruments includes lags of the endogenous variables and lags of the ir-mpr spread. 12 significantly positive, but does not entail a one-for-one change in the ir. In fact, the estimated coefficient in all the specifications lies in the range 0.15-0.22, which suggests that the effect on the ir of a one-time change in the mpr is distributed over time. This finding is also consistent with the evidence provided by Nautz and Offermanns (2006), Angelini (2002) and Linzert and Schmidt (2008). In Col [1] of Table 4 we test some calendar effects that could potentially affect the liquidity position of the banks: the day of the monetary policy meeting (which takes the value of one for contemporaneous and following day of the meeting), the value-added tax payment day (which takes the value of one on the day of payment and on the previous day) and the preceding four days to the end of the maintenance period.19 In addition, we include a dummy that takes the value of one in the first four days of the maintenance period in order to control for the higher demand because of the banks’ obligation to comply with at least 90% of the required reserve by the 23rd of each month. The aim of this intermediate target is to encourage less volatility in the banks’ reserve requirement compliance and thus the ir. Our results suggest that the ir does not vary significantly in the final days of the maintenance period. On the contrary, our results show that in the days prior to the VAT’s payoff day and on the days surrounding the monetary policy meeting, the ir increases approximately 2 bps, although in the case of the monetary policy meeting the increase is significant only in the three last specifications. Similarly, in each of the first four days of the maintenance period, the ir increases by 3 bps. This result is robust to different specifications of the length of dummy variables, and as we show in columns [2] to [4] of Table 4, they are also robust to the inclusion of alternative sets of control variables. 19 Consistently with former theoretical models which consider monthly reserve requirements as the most important shock affecting banks’ liquidity position (see Ho and Saunders, 1985; Freixas et al, 2000; Allen and Gale, 2000; King, 2004 among others), the end of maintenance period dummy variable is by far the most extensive calendar effect considered in the literature. See Hamilton (1996); Perez and Quiros (2002); Wurtz (2003); Prati et al. (2003) and Nautz and Offermanns (2006), among others. 13 Table 4 IV Estimations Dependent Variable: Δirt May 2005 – Aug 2008 [1] [2] [3] [4] ir (t-1) -0.342 *** -0.299 *** -0.295 *** -0.286 *** mpr (t-1) 0.343 *** 0.299 *** 0.295 *** 0.286 *** Δ mpr 0.215 ** 0.179 0.175 0.142 Δ ir (t-1) -0.092 *** 0.032 0.026 0.031 Δ mpr (t-1) 0.230 * 0.251 ** 0.261 ** 0.223 * Calendar Effects MP Meeting day 0.021 0.03 ** 0.032 ** 0.034 ** VAT Payoff day 0.026 ** 0.02 * 0.018 0.015 Pre- End of Maintenance Period -0.006 -0.004 -0.003 -0.003 Post- End of Maintenance Period 0.028 *** 0.03 *** 0.027 ** 0.028 ** Central Bank's Open Market Operations Δ Injections -0.057 *** -0.061 *** -0.062 *** Δ Injections (t-1) -0.003 -0.007 -0.007 Δ Draining 0.014 0.011 0.007 Δ Draininng (t-1) 0.056 ** 0.054 ** 0.059 ** Δ Mandatory Reserve Requirement 0.011 0.015 0.011 Δ Mandatory Reserve Requirement (t-1) 0.111 ** 0.112 *** 0.116 *** Δ Net Permanent Facilities 0.102 * 0.103 * 0.111 ** Δ Net Permanent Facilities (t-1) -0.091 -0.083 -0.095 [0.055] [0.057] [0.058] Private Investors Depositors Δ Other Investors' Deposits -0.008 -0.004 Δ Other Investors' Deposits (t-1) -0.006 -0.004 Δ Short-Run Pension Funds Deposits 0.091 0.093 Δ Short-Run Pension Funds Deposits (t-1) -0.112 * -0.123 * Δ Long-Run Pension Funds Deposits 0.110 0.118 Δ Long-Run Pension Funds Deposits (t-1) -0.142 * -0.141 * Other External & Domestic Variables Forward IR - MPR -0.043 Forward IR - MPR (t-1) -0.030 Libor-Ois 0.027 Libor-Ois (t-1) -0.022 VIX 0.017 VIX (t-1) -0.044 Observations 795 795 795 795 Note: * significant at 10%; ** significant at 5%; *** significant at 1% Source: Authors’ calculations. 14 The estimates considering monetary operations are presented in the block named “Central Bank’s Open Market Operations” in Cols. [2] to [4]. Both contemporaneous and lagged discretionary injections and drainages have the expected signs and similar magnitudes. The difference between injections and drainages lies in the timing of impact. While injections are statistically significant contemporaneously, drainages reach significance one period lagged (both coeffcients are statistically significant at the 95% confidence level).20 The estimated coefficients for these variables are robust along all the specifications and, more importantly, they are economically significant. However, the size of the average drainages and injection operations are quite different -$ 600 and $ 300 billion, respectively (Table 2). Therefore, in practice, when the central bank had operated through repos, the impact on the ir has been two times higher in draining than injecting liquidity. In effect, if we employ the average (median) size of each type of operation, the expected effect of drainages and injections on the ir is 3.4 (3.1) and 1.7 (1.4) bps, respectively. These results suggest that taking the ir to the mpr by the Central Bank is equally effective when the market has liquidity shortage with respect to liquidity abundance in the interbank system, but the average intervention implies that drainages are economically more significant than liquidity injections. Regarding permanent monetary operations, we obtain a positive contemporaneous effect, and a negative lagged effect of similar magnitude that offset the initial positive effect on the ir. In fact, it is not possible to reject the null hypothesis that the sum of both coefficients is statistically equal to zero at usual confidence intervals. In other words, the estimates imply that “changes” in the volume of the operations through permanent facilities do no affect permanently the ir. However, this does not imply that existence of this mechanism does not affect the dynamics of the ir. The reason behind this clarification is that this instrument is available every day and, therefore, the market could have internalized its operation in the valuation of the ir and, therefore, the use (or not) of this facility is already incorporated in the ir.21 Finally, the lagged change in the reserve requirement—which is proportional to 20 Hereafter we will refer to discretional injections and draining simply as “injections” and “draining”. 21 To test formally if the existence of this mechanism affects the dynamics of the ir, we should have data covering a period without the operation of the permanent facilities, which is not available. However, such analysis goes beyond the purpose of this paper. 15 the capital of each bank—is strongly significant, suggesting that higher reserve requirements reduce the liquidity of the banks and, therefore, increase the ir. In Cols. [3] and [4] we test the relevance of private deposit variables which supposedly provide liquidity to the banking system. We test the effect of deposits by splitting overall deposits into three categories: deposits by PF with duration of up to 90 days (short-run deposits), deposits by PF with duration longer than 90 days (long-run deposits) and deposits by other private investors. The only statistically significant variables are the lagged short- and long-run PF deposits.22 In order to evaluate the economic significance of long-run PF deposits, we employ the average daily change of short- long-run PF deposits — -$0.53 and $4.15 billion respectively—, leading to a potential effect on the ir reaching -0.006 and 0.05 bps. This magnitude is quite low. However, the potential effects of PF deposits could be quite important. In fact, if PF liquidate 25% of their short- and long-run deposits, the ir could go up by 30 bps (5.6 and 24.3 bps respectively).23 The final variables set we consider consists of mpr surprises (proxied by the forward ir – mpr spread) and two variables that capture the external environment: the Libor-OIS spread and the VIX. Neither of these variables prove to be statistical significant. In sum, we have two sets of candidates that correlate significantly with changes of the ir: standard calendar effects and central bank’s open market operations. In order to asses the relative statistical significance of those variables, in Table 5 we test the null hypothesis that all the regressors within a given variable set are non-significant for each specification. Results in the first row of Table 5 show lagged levels and differences of ir and mpr being strongly significant at the 1% level. Similarly, results in rows two and three confirm both calendar effects and central bank’s operations being statistical significant at the 1% level. Rows four and five reveal that private investors deposit variables and other external and domestic variables are both non-significant at standard confidence levels. We also compute three additional measures for the fit of each regression: the Akaike and the Bayesian Information Criterions (AIC and BIC, 22 We also consider deposits by private investors’ variables overall deposits and overall PF deposits (both short- and long-run). The results are qualitatively similar and are therefore not reported. 23 Notice that this estimation assumes perfect linearity -constant coefficient-, which could be a conservative assumption for this type of estimation. 16 respectively), and the regression’s adjusted R-Squared. AIC suggests that the model including including monetary operations and private investors deposits as preferred to specification 1 and 2, while BIC points to a specification only the calendar effects better adjusts the data (specification 1). On the contrary, specifications including other external variables (specifications 3 and 4, respectively) are never preferred by AIC or BIC to the more parsimonious specifications 1 and 2. Finally, the last row in Table 5 confirms that the inclusion of monetary operations into the model improves the adjusted R- Squared from 0.30 to 0.34, while the inclusion of further variables is unable to improve significantly the fit of the regression. Table 5 Wald Tests Dependent Variable: Δirt May 2005 – Aug 2008 [1] [2] [3] [4] Lagged Levels and Differences of ir and mpr 105.97 *** 67.22 *** 68.02 *** 61.43 *** Calendar Effects 19.41 *** 19.59 *** 16.67 *** 16.98 *** Central Bank's Open Market Operations 70.32 *** 73.34 *** 70.38 *** Private Investors Depositors 5.21 5.31 Other External & Domestic Variables 1.59 Observations 795 795 795 795 Akaike Information Criterion -2041.7 -2066.4 -2074.4 -2047.1 Bayesian Information Criterion -1999.6 -1986.8 -1966.8 -1911.4 R-squared 0.30 0.34 0.35 0.34 Note: * significant at 10%; ** significant at 5%; *** significant at 1% Source: Authors’ calculations. 3.1.2. Asymmetric Effects In this subsection we explore whether our estimates are sensitive to aggregate market liquidity, i.e. to positive and negative values of the ir - mpr spread. For such a purpose, we consider the specification excluding other domestic and external variables (Col. [3], Table 4).24 Operationally, we estimate a regression in which the speed of convergence, the contemporaneous and lagged changes in the ir and in the mpr, discretional monetary operations and deposits 24 The results are robust to different specifications of the ir dynamics. 17 variables are interacted with a dummy variable that takes the value one if the spread is positive and 0 otherwise. In order to determine whether the coefficients are statistically different in both cases, we compute Wald tests. Table 6 presents the coefficients for negative and positive spreads in Cols. [1] and [2] respectively, while Col. [3] shows the statistics of the Wald test under the null hypothesis of non-asymmetric coefficients –i.e., that the coefficients are not statistically different. The results reveal asymmetric effects on just some of the variables. The first source of asymmetry emerges from the lagged ir – mpr spread, which is significantly higher when the market is illiquid than with a liquid market (0.46 vs. 0.32). Secondly, results reveal the change in mpr being also asymmetric. In fact, when the spread is positive, changes in the mpr are translated into a one-to-one basis to the ir, but the pass-through is significantly lower if the spread is negative (less than one-fifth). Finally, we also find some evidence of asymmetries in mandatory reserves. Table 6 Testing Asymmetric Coefficients [1] [2] [3] Illiquid Market Liquid Market Chi2 ir > mpr ir < mpr ir (t-1) 0.462 *** 0.324 *** [2.31] mpr (t-1) -0.455 *** -0.329 *** [2.80] * Δ mpr 1.050 *** 0.130 [19.89] *** Δ ir (t-1) -0.086 -0.067 [0.07] Δ mpr (t-1) 0.488 *** 0.198 [1.43] Central Bank's Open Market Operations Δ Injections -0.041 *** -0.070 ** [0.66] Δ Injections (t-1) -0.017 -0.068 [1.18] Δ Draining -0.024 0.015 [2.00] Δ Draining (t-1) -0.006 -0.049 [3.25] * Δ Mandatory Bank Reserve Position 0.027 -0.050 [5.48] ** Δ Mandatory Bank Reserve Position (t-1) 0.067 ** -0.023 [8.52] *** Private Investors Depositors Δ Other Investors' Deposits -0.025 -0.078 [1.93] Δ Other Investors' Deposits (t-1) -0.022 0.017 [0.76] Δ Short-Run PF Deposits 0.118 0.175 [0.35] Δ Short-Run PF Deposits (t-1) -0.151 ** 0.026 [0.09] Δ Long- Run PF Deposits 0.000 0.068 [0.18] Δ Long- Run PF Deposits (t-1) 0.061 0.023 [1.00] Observations 680 Note: * significant at 10%; ** significant at 5%; *** significant at 1% Source: Authors’ calculations 18 3.1.3 Disentangling the Determinants of the ir: August 2007 In this section, based on the fourth specification in Table 4 (Col. [4]), we decompose the incidence of each one of the considered variables in the dynamics of the ir at the end of August 2007. This exercise is relevant because during this period the ir was systematically above the mpr—averaging 10.5 bps between August 10th and September 12th, with a peak of 21 bps—(Figure 1) and, additionally, it is not clear what were the main drivers behind this dynamic, i.e., PF deposits, international turmoil, expectations on the next monetary policy meeting, etc. Figure 3 presents the performance of the model in terms of explaining the ir (panel a) and, complementary, the contribution to the ir dynamics of the different explanatory variables (panel b). In general terms, the model has a relatively good performance at the beginning of August—recall that the model has a daily frequency—but it does a poor job between the end of August and the monetary policy meeting of September 13th. In particular, during the second week of September, the explanatory variables are able to explain just a minor part of the ir decoupling. In some sense, this behavior is consistent with the mix of uncertainty associated with the response of monetary policy to both inflationary pressures from the international financial turbulences that were affecting the economy at that moment. With respect to the explained ir dynamics—and particularly at the beginning of August— it is interesting to note that the calendar effects and monetary operations have an active role in the behavior of the ir. The variance decomposition indicates that the calendar effects, open market operations, private deposits and other controls account for 40, 56, 4 and 1% of the explained variance during the period under study, respectively. Moreover, the maximum contribution of the open market operations in the August-September episode is 11.1 bps when the difference between ir and mpr is close to 18 bps – the fifth of September. As for PF, their role is restricted to just a couple of days in mid-August that correspond to the loosening of the restrictions on PF to hold foreign assets.25 In fact, the estimations indicate that the maximum PF contribution to the ir occurred in the August episode, accounting for 2.9 bps of the gap between the ir and the mpr in the period. This point contrasts with 25 The limit to holding foreign assets was increased from 30% to 35% of total portfolio on August, 9th, 2007. 19 several opinions by market operators during those days regarding the impact on the market liquidity conditions due to potential PF’ portfolio adjustments.26 Figure 3 Explaining the ir Dynamics: August 2007 Panel (a) Panel (b) Explained versus Unexplained Share Decomposition of Model’s Explained Share Model Effective Monetary Operations PF Deposits 0.30 0.25 Calendar Effects Other Deposits Other Controls 0.20 0.20 0.15 0.10 0.10 0.05 0.00 0.00 -0.10 -0.05 -0.20 -0.10 1/8 8/8 16/8 23/8 30/8 6/9 13/9 1/8 8/8 16/8 23/8 30/8 6/9 13/9 Source: Authors’ calculations. 3.2. Panel Data Analysis This section presents panel data estimates with the purpose of exploring whether bank level differences could be relevant for explaining the dynamics of the ir. Due to data availability, the sample period covers a shorter period than in the previous section (from June 2006 to August 2008). The empirical model estimated in this section is similar to the model used in the time series section. In fact, we estimate a panel error correction model with IV to control for endogeneity of some explanatory variables:27 Δirit = α 0 ( irit −1 − mprt −1 ) + δ ' X i ,t + υi + ηt + ε it (2) where the dependent variable is defined as the change in the interbank asking rate defined in section 2.1, and where υι and ητ correspond to fixed and time effects, respectively,28 while the index i denotes each of the 29 banks in our sample. Vector X contains the same controls used for time series estimations with the only difference that they are disaggregated at the bank level. In general, the notation for the other variables remains the same as before. 26 See section 2. 27 A quite similar econometric approach is performed by Cocco et al. (2009) 28 We consider time effect with a weekly frequency. The reason behind this decision lies in the fact that the time dimension (539) of our dataset is significantly higher than the number of individuals (29). 20 Similarly to time series estimations, we apply the Hausman (1978) test to each covariate. The test supports the statistical exogeneity assumption for all the variables except for net permanent facilities. For this variable, we consider as external instruments the contemporaneous and lagged values of derivative contracts. The logic behind the use of derivatives as instruments is that they reduce the availability of banks to lend and borrow in the interbank market because they employ part of pre-determined credit lines between banks to operate between them, but the use of derivatives is not directly related to the ir dynamics. Therefore, the use of derivatives is related to the endogenous explanatory variable and, simultaneously, not to the error term. 3.2.1. Baseline Estimations Table 7 presents the benchmark regressions. The main difference with respect to time series estimations is the magnitude of the coefficients, which tend to be higher in the panel dimension. Concretely, the speed of convergence of the ir of each bank to the mpr is practically twice the speed at the aggregated level, roughly 0.50 (Table 7) versus 0.29 (Table 4). On the other hand, and perhaps more interestingly, the contemporaneous effect of injections fluctuates around 0.58, while in the time series estimates are close to 0.05 (however, this coefficient has low statistical significance). Similarly, the magnitude of the drainage effect is 10 times the effect estimated in the time series: 0.7 in panel estimations while in the time series it is approximately 0.06. A similar situation occurs with the estimated effect of injections, which is approximately three times higher than the estimated effect in the time series dimension. With respect to calendar effects, they have similar magnitudes compared to the time series estimates. The only difference is the estimated effect for the monetary policy meeting dummy variable, which turns out to be negative and significant, and the value-added payoff day dummy, which turns out to be non-significant. The set of calendar effects also include a dummy that takes the value one for those banks that are net lenders in each trading day. We do this in order to capture the fact that lender banks have probably more liquidity at hand and, therefore, should face a lower asking ir. Our belief is confirmed by the finding of an ir 0.7 bps lower for those banks. The existence of higher impacts at individual than aggregated level reflects the high heterogeneity across banks in terms of the use of instruments 21 and liquidity positions at each moment.29 In this sense, this result highlights the importance of a monetary planning that takes into account the liquidity position of each bank in order to maximize the efficiency of its instruments. To illustrate the implications of the magnitude of the individual elasticities, if the Central Bank reduces liquidity by $ 700 billion through liquidity deposits in 9 commercial banks,30 the ir of the banks using such instrument will go up by 5.3 bps. In contrast, time series estimates indicate that the aggregated ir will decrease by only 3.7 bps. In contrast, a liquidity injection of 300 billion through repo operations in 5.1 commercial banks reduce the ir of those banks in a similar amount than the estimated response in the time series section.31 Table 7 Panel IV Estimation. Dependent Variable: Δirit Sample Period: June 2006 to August 2008 [ 1 ] [ 2 ] [ 3 ] ir (t-1) - mpr (t-1) -0.542 *** -0.481 *** -0.479 *** Δ mpr 0.228 *** 0.436 *** 0.437 *** Δ ir (t-1) -0.058 *** -0.058 ** -0.059 ** Δ mpr (t-1) 0.161 *** 0.417 *** 0.417 *** Calendar Effects MP Meeting day -0.010 * -0.018 *** -0.017 *** VAT Payoff day -0.004 -0.001 -0.001 Pre- End of Maintenance Period -0.009 *** -0.007 * -0.007 * Post- End of Maintenance Period 0.026 *** 0.027 *** 0.027 *** Dummy Lender Bank -0.006 ** -0.007 ** -0.007 ** Central Bank's Open Market Operations Δ Injections -0.184 ** -0.181 ** Δ Injections (t-1) 0.13 0.135 Δ Drainage 0.725 ** 0.702 ** Δ Drainage (t-1) 0.679 0.649 Δ Net Permanent Facilities -0.541 -0.529 Δ Net Permanent Facilities (t-1) 0.151 0.219 Private Investors Depositors Δ Overall PFs Deposits -0.095 Δ Overall PFs Deposits (t-1) -0.294 Δ Other Deposits 0.005 Δ Other Deposits (t-1) 0.059 Observations 3804 3085 3085 Groups 22 20 20 Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. Source: Authors’ calculations. 29 If banks were not heterogeneous, the aggregate elasticity will collapse to the individual estimates. 30 The average liquidity deposit implies that 9.5 banks use this facility, each one requesting 73.1 billions pesos. 31 The average repo operations implies that 5.1 banks use this facility, each one requesting 61.7 billions pesos. 22 3.2.2. Asymmetric Response In this section we explore whether the asymmetric effects of monetary operations and deposits on the ir dynamics found in time series analysis are also valid on the ir of individual banks. Table 8 reports the coefficients for these variables depending on the sign of the ir – mpr spread of each bank—i.e., liquid and illiquid market. The results are similar to the aggregated estimates, but they also add some new pieces of information with respect to the interbank market dynamics. Table 8 Testing Asymmetry Dependent Variable: Δirit Sample Period: June 2006 to August 2008 [1] [2] [3] Illiquid Market Liquid Market ir > mpr ir < mpr Chi2 ir (t-1) - mpr (t-1) -0.547 *** -0.450 *** [207.4] *** Δ mpr 0.982 *** 0.074 * [127.1] *** Δ ir (t-1) 0.001 -0.106 *** [4.90] ** Δ mpr (t-1) 0.083 0.163 * [4.52] ** Central Bank's Open Market Operations Δ Injections -0.096 0.000 [0.19] Δ Injections (t-1) 0.125 -0.119 [0.00] Δ Drainage 0.088 0.765 *** [2.97] * Δ Drainage (t-1) 0.065 1.709 *** [4.93] ** Private Investors Depositors Δ Overall PFs Deposits 0.039 0.400 [0.62] Δ Overall PFs Deposits (t-1) -0.442 -0.380 [1.55] Δ Other Deposits 0.067 -0.267 [0.91] Δ Other Deposits (t-1) 0.120 ** -0.139 [0.01] Observations 3059 Groups 20 Notes: Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. The equations include the same controls used in the time series IV estimation Source: Authors’ calculations. 23 First, the speed of convergence is statistically higher when the banks are illiquid or, equivalently, the banks’ capacity to bring down the ir is higher when its ir is above the mpr. Moreover, and in line with time series estimations, the ir adjustment to changes in the mpr is statistically higher when the ir is above than below the mpr. In some sense, these results can be summarized as a higher bank capacity to adjust to an illiquid status—i.e., ir>mpr—and/or, complementary, the mpr tends to be the relevant marginal rate for illiquid banks, which is consistent with an ir higher than mpr, and vice versa. Regarding monetary operations, we find that discretionary drainages are more effective when the market is liquid. In fact, the magnitude of the sum of both contemporaneous and lagged draining coefficients when the market is liquid is almost ten times the coefficients when the market is illiquid. On the other hand, injections are not statistical significant. The same is true for PFs overall deposits, and for the contemporaneous value of depostis by other investors. 3.2.3 Large and Medium Banks In this subsection we evaluate if the responsiveness of the ir depends on the bank’s scale. For such purposes, and following the bank classification proposed by Jara and Oda (2007), we run a different regression for the large-, medium-, and small-banks. These authors make a cluster analysis for the Chilean banking industry, defining each cluster according to the Euclidean distance of each bank with respect to others which is dependent of a set of characteristics.32 Results in Table 8 reveal that large-scale banks are able to adjust faster to misalignments of the ir from the mpr. In addition, large-scale banks are less responsive to variations of the mpr. These results are consistent with the fact that large-sized banks have a bigger quantity of funding sources as well as greater assets. On the other extreme, small-scale banks shows the smaller coefficient of convergence and are more responsive to changes in the mpr. The main result, however, corresponds to the strong asymmetry observed in the monetary operations depending on bank type. Open market operations are non-significant for large- and medium-scale banks. This finding is consistent 32 Jara and Oda (2007) consider the following characteristics: market share, leverage, degree of portfolio diversification and target market. 24 with a higher degree of autonomy in the funding of this type of banks (see section 2.3). On the contrary, small-scale banks are the most responsive classification to open market operations. Actually, they are the only group responding significantly to both liquidity and draining operations. Table 9 Panel IV Estimation by Type of Bank Dependent Variable: Δirit Sample Period: June 2006 to August 2008 Large Banks Medium Banks Small Banks [1] [2] [3] ir (t-1) - mpr (t-1) -0.649 *** -0.55 *** -0.486 *** Δ mpr 0.134 0.197 *** 0.299 *** Δ ir (t-1) 0.002 -0.047 -0.083 *** Δ mpr (t-1) 0.185 0.08 0.21 ** Central Bank's Open Market Operations Δ Injections -0.025 -0.163 -0.266 * Δ Injections (t-1) 0.074 0.139 0.212 * Δ Drainage 0.294 0.687 1.36 *** Δ Drainage (t-1) -0.439 0.239 0.986 Private Investors Depositors Δ Overall PFs Deposits -0.224 0.338 0.932 Δ Overall PFs Deposits (t-1) -0.347 -0.233 -0.189 Δ Other Deposits -0.053 0.042 0.173 Δ Other Deposits (t-1) 0.049 0.012 0.047 Observations 497 1650 1395 Groups 4 9 7 Notes: Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. The equations include the same controls used in the time series IV estimation Source: Authors` calculations. 4. Conclusions In this paper we make use of three detailed micro-datasets to understand the determinants of the dynamics of the ir in Chile, which allows us to evaluate in detail the role of open market operations and private deposits. Regarding monetary policy, we consider discretionary operations that provide liquidity (repos) and those that reduce it (liquidity deposits), and non-discretionary 25 instruments (permanent credit facilities). Also, the estimations control for calendar effects, which are quite relevant in explaining the dynamics of the ir. In general terms, the main findings show the effectiveness of open market operations in terms to get the ir closer to the mpr. This is especially valid to discretionary operations. In fact, the point estimates of the coefficients related to drainages and injections are statistically and economically significant. Indeed, the average draining operation increases the ir approximately 3.4 bps, while the opposite operation reduces the ir around 1.7 bps. The asymmetries detected suggest that such effectiveness depends on the liquidity status at market level. Specifically, open market operations seem to be more effective when the market is illiquid –i.e., ir>mpr. Moreover, in this line of results, the estimates suggest that the pass trough from mpr to ir is close to 1 when the ir is above the mpr. Conversely, if the ir is below the mpr, the coefficient of pass trough is near to 0.15. The role of asymmetries of monetary operations is reinforced by the panel estimates. In general, the panel’s point estimates of open market operations are significant higher than the time series estimates –for instance, the magnitude of the drainage effect is also 10 times the effect estimated in the time series– which indicates that the effectiveness and access to central bank’s instruments is quite heterogeneous across banks. The estimates by category of bank -large and medium- show that part of this heterogeneity is captured trough this classification, where the open market operations are less relevant to large banks (the more liquid ones), and more relevant for small-sized banks. The results on PF deposits indicate that they are statistically significant with a relative high coefficient –specially, long-run deposits- but if we consider the behavior of this variable during the sample period, their economic relevance is limited. In fact, the statistical tests oriented to evaluate the relevance of potential explanatory variables suggest that bank’s deposits do not significantly help to improve the econometric specifications of the ir. Traditional information criteria statistics tilt towards a specification based on calendar effects and open market operations as controls. Nonetheless, from a financial stability perspective, the PF deposits could be quite relevant on the dynamics of the ir, because PF accounts for approximately 20% of total bank deposits. In other words, even though the PF’s deposits did not play an important role on the dynamics of the ir during the sample period, if these investors rebalance their 26 portfolios abruptly against banks’ deposits, the effects on the interbank market could have a systemic impact. Finally, the calendar effects are both statistical and economic significant. For instance, the day of payment of the value-added-tax is associated with an increase of the ir equal to 2 bps. This kind of result is relatively standard in the literature, and even though market practitioners could be habituated to them, they are puzzling. On one hand, these calendar effects are totally predictable - for instance, they are not doubts about when the taxes are paid- and, if we assume perfect markets, the ir should internalize such effects on its pricing. This line of reasoning opens important questions about the frictions that could be behind the calendar effects’ incidence. References Allen, F. and D. Gale (2000). “Asset Price Bubbles and Monetary Policy.” Working Papers 01-26, Center for Financial Institutions, Wharton School Center for Financial Institutions, University of Pennsylvania. Angenili, P. (2002). “Liquidity Announcement Effects in the Euro Area.” Paper 451, Banca D`Italia. Balduzzi, P., G. Bertola, and S. Foresi (1997). “A Model of Target Changes and the Term Structure of Interest Rates.” Journal of Monetary Economics 39, 223-49. Bartolini, L., B. Giusepe, and A. Prati (2002). “Day-to-Day Monetary Policy and the Volatility of the Federal Funds Interest Rate.” Journal of Money, Credit and Banking 34, 137-159. Balduzzi, P., G. Bertola, S. Foresi, and L. Klapper (1998). “Interest Rate Targeting and the Dynamics of Short-term Rates.” Journal of Money, Credit, and Banking 30, 26-50. Cocco J., F. Gomes, and N. Martins (2009). “Lending Relationships in the Interbank Market.” Journal of Financial Intermediation 18, 24-48. Cowan, K. D. Rappaport and J. Selaive (2006). “High Frequency Dynamics of the Nominal Exchange Rate in Chile.” Working Paper 433, Central Bank of Chile. Elliott, G., T. Rothenberg, and J. Stock (1996). “Efficient tests for an autoregressive unit root.” Econometrica 64, 813-836. 27 Freixas, X. B. Parigi and J-C. Rochet (2000). “Systematic Risk, Interbank Relations, and Liquidity Provisions by The Central Bank.” Journal of Money, Credit and Banking 32(3), 611-638 Gaspar, V. G. Quiroz and H. Medizabal (2004). “Interest Rate Determination in the Interbank Market.” Working Paper 351. European Central Bank. Hamilton J. (1996). “The Daily Market for Fed Funds”. Journal of Political Economy, 26-56. Hamilton, J. and O. Jorda (2002). “A Model For The Federal Fund Target” Department of Economics 99-07, California Davis - Department of Economics. Hausmann, J. A. (1978). “Specification Tests in Econometrics.” Econometrica 46, 1251 - 1271 Ho, T. and A. Saunders (1985). “A Micro Model of the Federal Funds Market.” The Journal of Finance 3, 977-988. Jara A. and D. Oda (2007). “¿Cómo Agrupar Las Instituciones Bancarias?: Un Aplicación de Análisis de Clusters.” Informe de Estabilidad Financiera, Primer Semestre 2007, Banco Central de Chile. King, T. (2004). “Discipline and Liquidity in the Market for Federal Funds.” Working Paper 2003-02, Federal Reserve Bank of St. Louis. Linzert and Schmidt (2008). “What Explains the Spread Between the Euro Overnight Spread and the ECB´s Policy Rate?”. Working Paper 983, European Central Bank. Nautz D. and C. Offermanns (2006). “The Dynamics Relationship between the Euro Overnight Rate, the ECB’s Policy Rate and the Term Spread.” Discussion Paper, Series 1: Economic Studies, 01/2006. Deutsche Bundesbank. Prati A., L. Bartolini and G. Bertola (2003). “The Overnight Interbank Market: Evidence from G-7 and the Euro Zone.” Journal of Banking and Finance, 27, 2045-2083. Raddatz, C. and S. Schmukler (2008). “Pension Funds and Capital Market Development: How Much Bang for the Buck?” Manuscript, World Bank. Sarno L. and D. Thornton (2002). “The Dynamics relationship Between the Federal Funds Rate and the Treasury Bill Rate: An Empirical Investigation.” Working Paper, 2000-032C. Federal Reserve Bank of St. Louis. 28 Spindt, P., and R. J. Hofmeister (1988). “The Micromechanics of the Federal Funds Market: Implications for Day-of-the-Week Effects in Funds Rate Variability.” Journal of Financial and Quantitative Analysis 23(4), 401- 416. Taylor, J. and J. Williams (2008). “A Black Swan in the Money Market.” Manuscript. Appendix 1: Managing the ir in Chile The Central Bank applies its monetary policy through the definition of a target level for the interbank rate (ir) known as the monetary policy rate (mpr). To ensure that the ir remains close to the mpr, the Central Bank must regulate the financial system’s liquidity (or reserves) through the use of several instruments: open market transactions, buying and selling of short-term promissory notes, lines of credit and liquidity deposits. Open market transactions are essentially carried out through regular auctions of promissory notes issued by the Central Bank: short-term nominal discount promissory notes (PDBC), and nominal and indexed promissory notes (BCP and BCU). Banks, financial institutions administering PF, insurance companies and mutual funds can participate in these tenders’ auctions. The bidding of promissory notes is carried out using the single price per auction method, that is, a cut-off rate is applied to all participants in the auction placing winning bids, in what is known as the “Dutch method”. This encourages competition among auction participants and tends to reflect current market conditions more accurately. In the case of the (average) ir deviates from the policy rate due, for instance as a result of liquidity levels below demand from the banking system, liquidity is injected to lower the ir rate and bring it closer to the mpr. This liquidity injection is generally achieved by overnight purchases of notes with repurchase clauses (repos). When the opposite occurs, and there is excess liquidity and the ir tends to be below the mpr, the excess is withdrawn by selling short term promissory notes. Additionally, starting on January 2005, the Central Bank implemented permanent credit (deposit) facilities which are intended to avoid the ir surpasses 29 (be below) the mpr by more (less) than 25 basis points. In this context, the implementation of open market operations followed common practices of developed economies’ central banks (US, Canada, Europe, among many others). The Central Bank uses permanent liquidity credit lines to provide financial institutions with overnight loans. This account requires collateral, instruments authorized in the Compendium of Financial Norms. It does not have quantitative limits, except for the availability of collateral of the applicant. Currently, the received interest rate is set at 25 basis points above the mpr. Similarly, permanent liquidity deposits allow financial institutions to deposit temporary excess liquidity overnight with the Central Bank and receive a minimum return. Currently, this rate is set at 25 basis points below the mpr and in practice this constitutes the floor of the ir. In order to regulate adequately financial system liquidity, the Central Bank develops a cash flow program around the reserve requirement time period, that is, from day nine of each month though day eight of the following month. To encourage less volatility in the banks’ reserve requirement compliance and thus the ir, there is also an intermediate reserve requirement on day 23 of each month, the deadline by which the banks must have complied with at least 90% of the required reserve. To program cash flow, projections are made for both supply and demand of bank reserves that is bills and coins in the power of banks and balances in banks’ current accounts in the Central Bank. Demand is of a derived nature that basically depends on reserve requirement rates and trends forecast for demand and term deposits, along with the behavior of currency in the public’s hands. The supply of bank reserves depends on the behavior of currency in the public’s hands and from the main sources of emission, particularly the maturities of previously auctioned promissory notes and other, more autonomous sources of monetary expansion for which projections are required. These operations include eventual purchases or sales of dollars within the financial system by the Central Bank and State financial operations having monetary effects. Once the supply and demand for bank reserves have been determined, the amount of notes to be tendered by the Central Bank is established. The calendar of auctions is published the day before each new reserve period begins. The liquidity projection for the next four weeks is monitored daily to permit 30 fine tuning operations on bank reserves, as needed, using the repo operations already mentioned or special sales of short-term promissory notes. Worth to mention that the mechanism to provide and dried out liquidity from banks described above is quite similar -with particularities in the implementation that may be crucial in the modeling strategy- in other economies. Appendix 2: Unit Root Test Table A1 Persistence, Unit Root and Co-integration Tests Interbank Target rate Rate Residual Half-Life Not Defined 345.2 1.3 Unit-Root-tests Augmented Dickey - Fuller -0.314 -0.409 -4.733 *** Phillips - Perron (Zt) -0.379 -0.863 -11.306 *** DF - GLS 1.675 0.31 -3.993 *** KPSS 7.95 *** 8.15 *** 0.884 *** Note: Except for KPSS, all the tests have as null hypothesis the non-stationarity of the series. * significant at 10%; ** significant at 5%; *** significant at 1% 31 Appendix 3: OLS Estimation (Baseline Time-Series Model) Table A2 OLS Estimations Dependent Variable: Δirt May 2005 – Aug 2008 [1] [2] [3] [4] ir (t-1) -0.343 *** -0.285 *** -0.281 *** -0.272 *** [0.040] [0.044] [0.044] [0.040] mpr (t-1) 0.343 *** 0.285 *** 0.281 *** 0.271 ** [0.041] [0.045] [0.044] [0.040] Δ mpr 0.217 ** 0.158 0.154 0.134 [0.099] [0.107] [0.109] [0.115] Δ ir (t-1) -0.094 *** -0.066 * -0.065 * -0.084 * [0.035] [0.037] [0.037] [0.036] Δ mpr (t-1) 0.232 * 0.268 * 0.277 * 0.233 [0.129] [0.142] [0.146] [0.149] Calendar Effects MP Meeting day 0.021 0.031 ** 0.032 ** 0.033 ** [0.014] [0.014] [0.014] [0.014] VAT Payoff day 0.025 ** 0.025 ** 0.023 ** 0.021 * [0.011] [0.011] [0.011] [0.012] Pre- End of Maintenance Period -0.007 -0.007 -0.005 -0.005 [0.006] [0.006] [0.006] [0.006] Post- End of Maintenance Period 0.027 *** 0.015 * 0.012 0.013 [0.009] [0.008] [0.009] [0.009] Central Bank's Open Market Operations Δ Injections -0.054 *** -0.058 *** -0.057 *** [0.015] [0.015] [0.015] Δ Injections (t-1) -0.005 -0.010 -0.011 [0.014] [0.014] [0.014] Δ Draining 0.017 0.014 0.013 [0.014] [0.014] [0.014] Δ Draininng (t-1) 0.006 0.006 0.005 [0.013] [0.012] [0.012] Δ Mandatory Reserve Requirement 0.014 0.017 0.018 [0.020] [0.019] [0.019] Δ Mandatory Reserve Requirement (t-1) 0.063 *** 0.067 *** 0.067 *** [0.019] [0.019] [0.019] Δ Net Permanent Facilities 0.110 *** 0.112 *** 0.113 *** [0.018] [0.018] [0.017] Δ Net Permanent Facilities (t-1) 0.033 * 0.033 * 0.036 ** [0.018] [0.018] [0.017] 32 Table A2 (cont.) OLS Estimations Dependent Variable: Δirt May 2005 – Aug 2008 [1] [2] [3] [4] Private Investors Depositors Δ Other Investors' Deposits -0.018 -0.014 [0.019] [0.019] Δ Other Investors' Deposits (t-1) -0.007 -0.005 [0.015] [0.015] Δ Short-Run Pension Funds Deposits 0.089 0.084 [0.067] [0.067] Δ Short-Run Pension Funds Deposits (t-1) -0.051 -0.057 [0.051] [0.051] Δ Long-Run Pension Funds Deposits 0.107 0.109 [0.081] [0.080] Δ Long-Run Pension Funds Deposits (t-1) -0.125 ** -0.119 * [0.062] [0.062] Other External & Domestic Variables Forward IR - MPR -0.016 [0.052] Forward IR - MPR (t-1) -0.067 [0.056] Libor-Ois 0.087 [0.160] Libor-Ois (t-1) -0.108 [0.149] VIX 0.001 [0.052] VIX (t-1) 0.009 [0.052] Observations 804 804 804 804 Akaike Information Criterion -2068.3 -2184.6 -2181.6 -2178.5 Bayesian Information Criterion -2021.4 -2100.2 -2069.1 -2037.8 R-squared 0.3 0.41 0.42 0.42 Robust standard errors in brackets * significant at 10%; ** significant at 5%; *** significant at 1% 33 Appendix 4: Full Tables Table A3 (Extended Table 4) IV Estimations; Dependent Variable: Δirt Sample Period: May 2005 – Aug 2008 [1] [2] [3] [4] ir (t-1) -0.342 *** -0.299 *** -0.295 *** -0.286 *** [0.041] [0.045] [0.043] [0.042] mpr (t-1) 0.343 *** 0.299 *** 0.295 *** 0.286 *** [0.041] [0.045] [0.043] [0.043] Δ mpr 0.215 ** 0.179 0.175 0.142 [0.099] [0.109] [0.109] [0.116] Δ ir (t-1) -0.092 *** 0.032 0.026 0.031 [0.035] [0.065] [0.067] [0.069] Δ mpr (t-1) 0.230 * 0.251 ** 0.261 ** 0.223 * [0.129] [0.119] [0.123] [0.127] Calendar Effects MP Meeting day 0.021 0.03 ** 0.032 ** 0.034 ** [0.014] [0.015] [0.015] [0.015] VAT Payoff day 0.026 ** 0.02 * 0.018 0.015 [0.012] [0.012] [0.012] [0.013] Pre- End of Maintenance Period -0.006 -0.004 -0.003 -0.003 [0.006] [0.006] [0.006] [0.006] Post- End of Maintenance Period 0.028 *** 0.03 *** 0.027 ** 0.028 ** [0.009] [0.011] [0.011] [0.011] Central Bank's Open Market Operations Δ Injections -0.057 *** -0.061 *** -0.062 *** [0.017] [0.017] [0.017] Δ Injections (t-1) -0.003 -0.007 -0.007 [0.014] [0.014] [0.014] Δ Draining 0.014 0.011 0.007 [0.030] [0.031] [0.031] Δ Draininng (t-1) 0.056 ** 0.054 ** 0.059 ** [0.026] [0.026] [0.026] Δ Mandatory Reserve Requirement 0.011 0.015 0.011 [0.030] [0.029] [0.029] Δ Mandatory Reserve Requirement (t-1) 0.111 ** 0.112 *** 0.116 *** [0.028] [0.028] [0.028] Δ Net Permanent Facilities 0.102 * 0.103 * 0.111 ** [0.058] [0.057] [0.056] Δ Net Permanent Facilities (t-1) -0.091 -0.083 -0.095 [0.055] [0.057] [0.058] 34 Table A3 (cont.) IV Estimations; Dependent Variable: Δirt Sample Period: May 2005 – Aug 2008 [1] [2] [3] [4] Private Investors Depositors Δ Other Investors' Deposits -0.008 -0.004 [0.020] [0.020] Δ Other Investors' Deposits (t-1) -0.006 -0.004 [0.017] [0.017] Δ Short-Run Pension Funds Deposits 0.091 0.093 [0.078] [0.078] Δ Short-Run Pension Funds Deposits (t-1) -0.112 * -0.123 * [0.068] [0.069] Δ Long-Run Pension Funds Deposits 0.110 0.118 [0.088] [0.089] Δ Long-Run Pension Funds Deposits (t-1) -0.142 * -0.141 * [0.079] [0.080] Other External & Domestic Variables Forward IR - MPR -0.043 [0.048] Forward IR - MPR (t-1) -0.030 [0.042] Libor-Ois 0.027 [0.058] Libor-Ois (t-1) -0.022 [0.059] VIX 0.017 [0.181] VIX (t-1) -0.044 [0.177] Observations 795 795 795 795 Akaike Information Criterion -2037.7 -2069.4 -2062.9 -2055.6 Bayesian Information Criterion -1991.0 -1985.4 -1950.9 -1915.6 R-squared 0.33 0.37 0.37 0.37 Note: Robust standard error in brackets. * significant at 10%; ** significant at 5%; *** significant at 1% 35 Table A4 (Extended Table 6) Testing Asymmetric Coefficients Dependent Variable: Δirit ; Sample Period: June 2006 to August 2008 [1] [2] [3] Illiquid Market Liquid Market ir > mpr ir < mpr Chi2 ir (t-1) 0.462 *** 0.324 *** [2.31] [0.074] [0.063]*** mpr (t-1) -0.455 *** -0.329 *** [2.80] * [0.073] [0.064]*** Δ mpr 1.050 *** 0.130 [19.89] *** [0.191] [0.093] Δ ir (t-1) -0.086 -0.067 [0.07] [0.075] [0.083] Δ mpr (t-1) 0.488 *** 0.198 [1.43] [0.142] [0.191] Central Bank's Open Market Operations Δ Injections -0.041 *** -0.070 ** [0.66] [0.014] [0.035]** Δ Injections (t-1) -0.017 -0.068 [1.18] [0.015] [0.048] Δ Draining -0.024 0.015 [2.00] [0.028] [0.031] Δ Draining (t-1) -0.006 -0.049 [3.25] * [0.026] [0.032] Δ Mandatory Bank Reserve Position 0.027 -0.050 [5.48] ** [0.028] [0.034] Δ Mandatory Bank Reserve Position (t-1) 0.067 ** -0.023 [8.52] *** [0.028] [0.035] Private Investors Depositors Δ Other Investors' Deposits -0.025 -0.078 [1.93] [0.019] [0.034]** Δ Other Investors' Deposits (t-1) -0.022 0.017 [0.76] [0.017] [0.042] Δ Short-Run PF Deposits 0.118 0.175 [0.35] [0.078] [0.106] Δ Short-Run PF Deposits (t-1) -0.151 ** 0.026 [0.09] [0.068] [0.148] Δ Long- Run PF Deposits 0.000 0.068 [0.18] [0.068] [0.102] Δ Long- Run PF Deposits (t-1) 0.061 0.023 [1.00] [0.065] [0.119] Observations 680 Note: * significant at 10%; ** significant at 5%; *** significant at 1% 36 Table A5 (Extended Table 7) Panel IV Estimation, Dependent Variable: Δirit Sample Period: June 2006 to August 2008 [1] [2] [3] ir (t-1) - mpr (t-1) -0.542 *** -0.481 *** -0.479 *** [0.024] [0.024] [0.024] Δ mpr 0.228 *** 0.436 *** 0.437 *** [0.041] [0.069] [0.069] Δ ir (t-1) -0.058 *** -0.058 ** -0.059 ** [0.016] [0.027] [0.027] Δ mpr (t-1) 0.161 *** 0.417 *** 0.417 *** [0.060] [0.056] [0.056] Calendar Effects MP Meeting day -0.010 * -0.018 *** -0.017 *** [0.005] [0.006] [0.006] VAT Payoff day -0.004 -0.001 -0.001 [0.004] [0.005] [0.006] Pre- End of Maintenance Period -0.009 *** -0.007 * -0.007 * [0.003] [0.004] [0.003] Post- End of Maintenance Period 0.026 *** 0.027 *** 0.027 *** [0.004] [0.003] [0.004] Dummy Lender Bank -0.006 ** -0.007 ** -0.007 ** [0.003] [0.003] [0.003] Central Bank's Open Market Operations Δ Injections -0.184 ** -0.181 ** [0.084] [0.086] Δ Injections (t-1) 0.13 0.135 [0.086] [0.086] Δ Drainage 0.725 ** 0.702 ** [0.319] [0.328] Δ Drainage (t-1) 0.679 0.649 [0.501] [0.499] Δ Net Permanent Facilities -0.541 -0.529 [0.627] [0.630] Δ Net Permanent Facilities (t-1) 0.151 0.219 [0.915] [0.915] Private Investors Depositors Δ Overall PFs Deposits -0.095 [0.219] Δ Overall PFs Deposits (t-1) -0.294 [0.301] Δ Other Deposits 0.005 [0.045] Δ Other Deposits (t-1) 0.059 [0.044] Observations 3804 3085 3085 Groups 22 20 20 Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. 37 Table A6 (Extended Table 8) Testing Asymmetry; Dependent Variable: Δirit Sample Period: June 2006 to August 2008 [1] [2] [3] Illiquid Market Liquid Market ir > mpr ir < mpr Chi2 ir (t-1) - mpr (t-1) -0.547 *** -0.450 *** [207.4] *** [0.035] [0.060] Δ mpr 0.982 *** 0.074 * [127.1] *** [0.086] [0.038] Δ ir (t-1) 0.001 -0.106 *** [4.90] ** [0.030] [0.041] Δ mpr (t-1) 0.083 0.163 * [4.52] ** [0.083] [0.098] Central Bank's Open Market Operations Δ Injections -0.096 0.000 [0.19] [0.136] [0.139] Δ Injections (t-1) 0.125 -0.119 [0.00] [0.098] [0.179] Δ Drainage 0.088 0.765 *** [2.97] * [0.417] [0.283] Δ Drainage (t-1) 0.065 1.709 *** [4.93] ** [0.521] [0.569] Private Investors Depositors Δ Overall PFs Deposits 0.039 0.400 [0.62] [0.241] [0.440] Δ Overall PFs Deposits (t-1) -0.442 -0.380 [1.55] [0.394] [0.354] Δ Other Deposits 0.067 -0.267 [0.91] [0.047] [0.188] Δ Other Deposits (t-1) 0.120 ** -0.139 [0.01] [0.057] [0.192] Observations 3059 Groups 20 Notes: Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. The equations include the same controls used in the time series IV estimation 38 Table A7 (Extended Table 9) Panel IV Estimation by Type of Bank; Dependent Variable: Δirit ; Sample Period: June 2006 to August 2008 Large Banks Medium Banks Small Banks [1] [2] [3] ir (t-1) - mpr (t-1) -0.649 *** -0.55 *** -0.486 *** [0.080] [0.034] [0.038] Δ mpr 0.134 0.197 *** 0.299 *** [0.110] [0.060] [0.085] Δ ir (t-1) 0.002 -0.047 -0.083 *** [0.104] [0.034] [0.031] Δ mpr (t-1) 0.185 0.08 0.21 ** [0.241] [0.098] [0.103] Central Bank's Open Market Operations Δ Injections -0.025 -0.163 -0.266 * [0.182] [0.178] [0.150] Δ Injections (t-1) 0.074 0.139 0.212 * [0.132] [0.128] [0.119] Δ Drainage 0.294 0.687 1.36 *** [0.287] [0.445] [0.378] Δ Drainage (t-1) -0.439 0.239 0.986 [1.129] [0.368] [0.935] Private Investors Depositors Δ Overall PFs Deposits -0.224 0.338 0.932 [0.543] [0.238] [1.099] Δ Overall PFs Deposits (t-1) -0.347 -0.233 -0.189 [0.685] [0.214] [1.120] Δ Other Deposits -0.053 0.042 0.173 [0.110] [0.053] [0.156] Δ Other Deposits (t-1) 0.049 0.012 0.047 [0.111] [0.086] [0.145] Observations 497 1650 1395 Groups 4 9 7 Notes: Robust standard errors in brackets; * significant at 10%; ** significant at 5%; *** significant at 1%. Estimations include individual and monthly fixed effects. The equations include the same controls used in the time series IV estimation 39 Documentos de Trabajo Working Papers Banco Central de Chile Central Bank of Chile NÚMEROS ANTERIORES PAST ISSUES La serie de Documentos de Trabajo en versión PDF puede obtenerse gratis en la dirección electrónica: www.bcentral.cl/esp/estpub/estudios/dtbc. Existe la posibilidad de solicitar una copia impresa con un costo de $500 si es dentro de Chile y US$12 si es para fuera de Chile. Las solicitudes se pueden hacer por fax: (56-2) 6702231 o a través de correo electrónico: bcch@bcentral.cl. Working Papers in PDF format can be downloaded free of charge from: www.bcentral.cl/eng/stdpub/studies/workingpaper. Printed versions can be ordered individually for US$12 per copy (for orders inside Chile the charge is Ch$500.) Orders can be placed by fax: (56-2) 6702231 or e-mail: bcch@bcentral.cl. DTBC-515 Abril 2009 Sovereign Defaulters: Do International Capital Markets Punish Them? Miguel Fuentes y Diego Saravia DTBC-514 Abril 2009 En Búsqueda de un Buen Benchmark Predictivo para la Inflación Pablo Pincheira y Álvaro García DTBC-513 Marzo 2009 From Crisis to IMF-Supported Program: Does Democracy Impede the Speed Required by Financial Markets? Ashoka Mody y Diego Saravia DTBC-512 Diciembre 2008 A Systemic Approach to Money Demand Modeling Mauricio Calani, Rodrigo Fuentes y Klaus Schmidt-Hebbel DTBC-511 Diciembre 2008 Forecasting Inflation in Difficult Times Juan Díaz y Gustavo Leyva DTBC-510 Diciembre 2008 Overoptimism, Boom-Bust Cycles, and Monetary Policy in Small Open Economies Manuel Marfán, Juan Pablo Medina y Claudio Soto DTBC-509 Diciembre 2008 Monetary Policy Under Uncertainty and Learning: An Overview Klaus Schmidt-Hebbel y Carl E. Walsh DTBC-508 Diciembre 2008 Estimación de Var Bayesianos para la Economía Chilena Patricio Jaramillo DTBC-507 Diciembre 2008 Chile’s Growth and Development: Leadership, Policy-Making Process, Policies, and Results Klaus Schmidt-Hebbel DTBC-506 Diciembre 2008 Exit in Developing Countries: Economic Reforms and Plant Heterogeneity Roberto Álvarez y Sebastián Vergara DTBC-505 Diciembre 2008 Evolución De La Persistencia Inflacionaria En Chile Pablo Pincheira DTBC-504 Noviembre 2008 Robust Learning Stability with Operational Monetary Policy Rules George W. Evans y Seppo Honkapohja DTBC-503 Noviembre 2008 Riesgo de Crédito de la Banca Rodrigo Alfaro, Daniel Calvo y Daniel Oda DTBC-502 Octubre 2008 Determinacy, Learnability, And Plausibility In Monetary Policy Analysis: Additional Results Bennett T. McCallum DTBC-501 Octubre 2008 Expectations, Learning, And Monetary Policy: An Overview Of Recent Research George W. Evans y Seppo Honkapohja DTBC-500 Octubre 2008 Higher Order Properties of the Symmetrically Normalized Instrumental Variable Estimator Rodrigo Alfaro DTBC-499 Octubre 2008 Imperfect Knowledge And The Pitfalls Of Optimal Control Monetary Policy Athanasios Orphanides y John C. Williams

DOCUMENT INFO

Shared By:

Categories:

Tags:
interbank market, central bank, interbank rates, interbank rate, central banks, interest rate, open market operations, last week, money market, money markets, libor rates, counterparty risk, budgetary allocations, adverse selection, lending rates

Stats:

views: | 24 |

posted: | 8/30/2010 |

language: | Spanish |

pages: | 44 |

OTHER DOCS BY vgg76349

How are you planning on using Docstoc?
BUSINESS
PERSONAL

By registering with docstoc.com you agree to our
privacy policy and
terms of service, and to receive content and offer notifications.

Docstoc is the premier online destination to start and grow small businesses. It hosts the best quality and widest selection of professional documents (over 20 million) and resources including expert videos, articles and productivity tools to make every small business better.

Search or Browse for any specific document or resource you need for your business. Or explore our curated resources for Starting a Business, Growing a Business or for Professional Development.

Feel free to Contact Us with any questions you might have.